Objective: The objective of this study was to test whether effects of b-amyloid (Ab) pathology on episodic memory were mediated by metabolism and gray matter volume in the early stages of Alzheimer disease.Methods: This was a prospective cohort study. We measured baseline Ab (using florbetapir-PET), brain function (using fluorodeoxyglucose-PET), and brain structure (using MRI). A mediation analysis was performed to test whether statistical effects of Ab positivity on cross-sectional and longitudinal episodic memory were mediated by hypometabolism or regional gray matter volume in cognitively healthy controls (CN, n 5 280) and mild cognitive impairment (MCI, n 5 463).Results: Lower memory scores were associated with Ab positivity (CN, mildly; MCI, strongly), smaller gray matter volumes (CN, few regions, including hippocampus; MCI, widespread), and hypometabolism. Smaller volumes and hypometabolism mediated effects of Ab in MCI but not in CN. The strongest individual regions mediated up to approximately 25%. A combination of brain structure and function mediated up to approximately 40%. In several regions, gray matter atrophy and hypometabolism predicted episodic memory without being associated (at p , 0.05) with Ab positivity.Conclusions: Changes in brain structure and function appear to be, in part, downstream events from Ab pathology, ultimately resulting in episodic memory deficits. However, Ab pathology is also strongly related to memory deficits through mechanisms that are not quantified by these imaging measurements, and episodic memory decline is partly caused by Alzheimer disease-like brain changes independently of Ab pathology. Alzheimer disease (AD)-like changes in brain structure and function may be present in the absence of biomarker evidence of b-amyloid (Ab) pathology.1 Therefore, it would be useful to quantify to what extent effects of Ab on cognition are explained by brain structure and function and to quantify the strength of Ab-independent associations between cognition and brain injury biomarkers. In this study, we tested the hypotheses that (1) Ab, and brain structure and function were associated with episodic memory deficits, (2) effects of Ab on episodic memory were mediated by brain structure and function, and (3) brain structure and function had Ab-independent effects on memory. Previous studies examining the relationships among Ab, cognition, and brain structure and function have mainly analyzed hippocampus, the ventricles, or whole brain lobes in combined or small cohorts of healthy controls (CN), patients with mild cognitive impairment